An Introduction to Economic Models

Edward R. Morey: Draft January 24, 2011

Economic models are, in some but not all ways, much like fashion models, the good ones are stylized, useful, and attractive representations of reality.

Simply put, economists build and test models of economic systems, or build and test models of a part of an economic system

What is the purpose of a model (theory)?

A model tries to correctly explain and predict the working of a system, and, in particular how things will change if something exogenous to the system changes. Note that model and theory are two different names for the same thing: the former is simply not ostentatious. 1

The following are systems The U.S. economy The solar system A hot dog sitting in some cold water on your stove The human body A household The market for potatoes The flow and distribution of water in the Mississippi delta Wanda Sue is a system, so it everyone else. For example: the system might be the market for cigarettes and the intent of the model is to explain and predict what will happen to the consumption of cigarettes if the price of cigarettes increases by one dollar. In this case one is not trying to explain/model how the price of cigarettes is determined, but rather than how consumption of cigarettes is affected by the price of cigarettes. That is, the price influences the system one is trying to explain, but the price is not part of what is trying to explain. In the recent recession, the Federal Government gave a subsidy to first-time home buyers (now expired) to simulate the housing market. Some economists built models to predict how this subsidy would affect the number of homes purchased, both in the short run and long run. That is, a model to predict how many homes would be purchased in a world where the subsidy is zero and a world where the subsidy is the amount specified by the government. Some of those models predicted the subsidy would lead to an increase in home buying, while some of the other models predicted only that the subsidy would cause people to buy sooner rather than later. The models predicted different things because they made different assumptions.

That said, I would rather date a model than a theory.

Building a model
One begins building a model by identifying the system of interest (the system you what to explain) A model cannot explain everything, only a subset of everything, usually a model explains only a small subset of everything. The model describes the chosen system in terms of variables and the relationships between those variables.

Variables
Variables are simply thing that vary: your temperature, the number of cigarettes people in Boulder smoke per day, how many people get married each day, whether an individual will have sex tomorrow, GDP, and the unemployment rate.

Representing variables and their levels

We typically use different letters (e.g. x, y, and z) to represent different variables.

For example, in a model we might use c to denote the variable aggregate consumption in the U.S and c t to represent the level of aggregate consumption in year t, For example c 2007 would be the aggregate level of consumption in 2007

Or, in another model c might be the number of cigarettes an individual smokes in 24 hours. In which case c i would be the number smoked by individual i, i=1,2,3,..,N. Maybe d i =1 if the condemned prisoner i is dead, and zero otherwise. In this model, dead would be considered a success. Variables that can take only two values are called dichotomous or dummy variables: one only needs two numbers (0 and 1, or, for example, 1 and 2, or 1 and 37) to represent a variable that can take only two levels (alive or dead, true or false, yes or no).

Endogenous and exogenous variables

In models one makes a distinction between those variables whose levels you want to explain in your model, and those variables you want to include in your model, but not explain. The former are called endogenous variables, the latter exogenous variables.

In more detail: you choose what variables you want your model to explain. The variables whose levels you want to explain are call endogenous variables. The adjective endogenous means inside what is determined inside your model.

The prefix ex mean out, from, or away The prefix end means in, into, into The adjective exogenous means the variable is determined outside of the system The adjective endogenous means the variable is determined within the system You include exogenous variables in a model because you assume they will determine/explain the levels of the endogenous variables in your model.

Assumptions relate variables

You then make assumptions about relationships between the variables in your model with the intent of explaining the levels of the endogenous variables in your model in terms of the levels of the exogenous variables in your model.

The exogenous variables in your model are the variables that you think determine the levels of the endogenous variables in your model. In your model you dont give a shit about how the levels of the exogenous variables are determined; you only care about how they influence the levels of the endogenous variables. Variables that are exogenous in one model might be endogenous in another model

Variables that are endogenous in one model might be exogenous in another model. It is your model so you can make whatever you want exogenous or endogenous, a variable just cannot be both.

Examples of endogenous and exogenous variables in different models

You want to build a model to explain, in the U.S. economy, yearly GDP (gross domestic product), aggregate consumption, and the level of unemployment: you want the model to predict the levels of these three variables. You assume that that the levels of these variables are determined by interest rates, the level of government expenditures and whether the President is a Republican or Democrat. For the purpose of this model you are not trying to explain how interest rates are determined, or how government expenditures are determined, or who gets elected President. In this model, GDP, aggregate consumption and the level of unemployment are endogenous variables. In this model interest rates, government expenditures and the political party of the President are exogenous variables. Our next model will involve nose picking, so I am going to take a detour and read a poem about nose picking.

Poem by Paul Hughes 2008

The following is a poem, not a model.

A foolish boy named Jimmy Price refused to take his mums advice and kept on poking up his nose his fingers, thumbs and, once, his toes nose picking gave him endless joy oh what a frightful, horrid, boy his mother told him that his head would cave right in and hed be dead but Jimmy knew that this was bluff its such an awful lot of guff theres really nothing wrong with snot Ill pick my nose and eat the lot he scraped and plucked his nostrils clean it was a truly gruesome scene ignoring every sign of pain he rummaged til he found his brain which out he pulled without a thought and thats the end of my report for Jimmy died right there and then he never picked his nose again

Moral

Its most unsatisfactory to eat what is olfactory.

A model of relationship status (the endogenous variable) with nose picking being an exogenous variable You want to build a model to explain only whether someone has a significant other, and, for the purpose of your model, you are willing to assume one either does or does not have a significant other there are no intermediate cases (e.g. my local friend whose girlfriend lives with her husband in California) Let s=1 of the individual has a significant other, and s=0 if they do not. So s is an endogenous variable in this model, a dichotomous variable. You are not trying to explain ones income or nose-picking habits but assume they are important determinants of whether one has a significant other. In this model, significant other is the only endogenous variable, and income and nose-picking habit are exogenous variables. y i could denote the individual is income and n i =1 if the individual is a nosepicker You might assume that the probability of having a significant other increases with ones income and, ceteris paribus, decreases if one picks their nose. 2 The above is not a complete model; rather it is the skeleton/outline of a model. A different model: Instead of the last model, maybe you want to explain both whether an individual has a significant other and whether the individual is a nose picker, and are willing to assume both are determined by ones income. In this model, significant other and nose picking are both endogenous variables. Income is still an exogenous.

Alternatively, you want to build a model to explain how many days an individual will ski Vail this winter (my dissertation predicted ski trips).

You assume it depends on the price of a Vail lift ticket, this winters snowfall, and how far the individual lives from Vail. You are not trying to model how Vail sets their liftticket prices, or explain how God (or global warming) decides how much it will snow, or explain why people choose to live where they live. In this model, number of ski days is the endogenous variable; price, snow fall, etc. are exogenous variables.

Ceteris paribus means everything else constant. Or, you could assume the opposite, its your model.

The state of the human body (dead or alive) would be of interest in a model designed to explain and predict the outcome of attempted executions. Life status, dead or alive, would be the endogenous variable. Volts administered might be an exogenous variable.

A model consists of three parts: definitions, assumptions, and predictions

Synonyms for predictions are hypotheses and if then statements.

Definitions
One begins model building by choosing and defining the endogenous and exogenous variables in your model.

For example, a potato is ..

The price of a potato is the wholesale price at the Chicago commodities exchange.

A hot dog is ..

Assumptions

One specifies, by assumption, which variables are endogenous and which are exogenous.

For example, one might assume the price and quantity sold of potatoes is what you want to explain, so these two variables are assumed endogenous variables.

Alternatively you might specify rainfall and temperature as exogenous variables (something influential that you dont want to explain).

One then specifies, by assumption, relationships between variables.

For example, one might assume the relationship between rainfall and potato production is an inverted U (production is low if rainfall is low or high)

This is the graph of an assumption. The specific assumption is lbs =100-(1005t+.5t^{1.5}).

Or, in another theory, one might assume that aggregate consumption in the U.S. increases, linearly, when aggregate income increases. Or, in notation, C = a +bY. Where b>0

Note that c=10+.75(income) is a much more restrictive assumption than is c=a+b(income).

A third example of an assumption

Boys pick their noses twice as much as girls.

Note how an assumption in one model might be a prediction in another model Models have more than one assumption and the assumptions cannot contradict one another

Equilibrium is an important assumption is most models. Class exercise: Choose and define two or three variables that you might want to put in a model (whatever variables you want). Define, in words, each of your variables. Then choose a different letter to denote each of your variables. Then specify two assumptions relating your variables. Specify each of your assumptions in words, mathematical notation, and with a graph.

Predictions
The predictions of a model follow logically from the assumptions and definitions.

Consider the following two assumptions: All men cry George Bush is a man

What prediction follows? Another model: Consider the following five assumptions: 1. 2. 3. 4. 5. Sad men cry Barak Obama is a man Barak Obama is a Democrat Existence of the Tea Party makes Democrats sad The tea party exists.

Assumptions 4 and 5 imply that Democrats are sad. This and assumption 3 imply that Barak is sad This and 2 imply that Barak is a sad man. This and 1 imply that Barak is crying. What are some other things the theory implies and does not imply? For example: It implies that Democratic women are sad, but it does not imply that they are crying, maybe they are and maybe they are not.

Assumption: The average person in the U.S. weighs 180 pounds. Assumption: Individual x weight more than the average person

Prediction X weights more than 180 pounds

A model can have assumptions that are not true in a literal senseassumptions simply cannot contradict one another

If you build a model you can define variables however you want. And you can assume whatever you want, as long as you do not contradict yourself. For example, you cannot simultaneously assume the following three things: A is greater than B B is greater than C C is greater than A. (make sure you can explain why any two of these assumptions contradict the third one) If you were building a model to predict the shots of an expert pool player, you might assume in your model that expert pool players know advanced geometry. Such an assumption is probably not literally true, but it might lead to accurate pool-shot predictions.

And you have to follow the rules of logic when deriving the predictions implied by your definitions and assumptions.

What scientists do is build models, test models, or both

This is true of chemists, biologists, and even economistsyes economists are scientists.

In models, what goes in (the assumptions) determines what comes out (the predictions). Logic is the machine that links the two.

Theory is another word for model.

Economists build theories

How do we judge (test) our models?

We see how well they explain/predict the system of interest.

For example imagine a model built to explain the driving habits of Americans: whether one owns a car, if so, what kind, and if one owns a car how many miles one drives per week.

Imagine this model consists of a bunch of definitions and assumptions and predicts that every time the price of gas increases by 1%, miles driven decreases by .4%

One can look at data to see if this is true.

If a model gets it all wrong, it is a baaad model?

Baaad is a matter of degree.

For example Newtonian physics predicts a lot of stuff correctly but gets some stuff wrong.

It has been replaced by Relativity theory which predicts correctly what Newtonian physics predicts, but, in addition, gets correct stuff that Newtonian physics got wrong, or did not consider.

How does one fix/modify a model/theory?

One changes the assumptions.

Consider the prevailing economic theory, in 1929, of how the economy works. The stock market crashed October 29, 1929 (Black Tuesday) at the beginning of the Great Depression, a long period of high unemployment (25%), and low income that ended only after immense government intervention to stimulate the economy (first President Roosevelts New Deal and finally World War II).

Dorothea Lange's Migrant Mother depicts destitute pea pickers in California, centering on Florence Owens Thompson, age 32, a mother of seven children, in Nipomo, California, March 1936.

Back then economists believed that

left along they [markets] were self-correcting and would return to an equilibrium that efficiently utilized capital, workers and natural resources this was the inviolate and core axiom of scientific economics itself A month after the Great Crash, the economists of Harvard, stated that a severe depression like that of 1920-21 is outside of the range of probability. (Richard Parker, John Kenneth Galbraith: his life, politics and economics, 2005, p.12) 3

They could not have been more wrong. A new theory emerged, Keynesianism, with the publication of John Maynard Keynes The General Theory of Employment, Interest and Money. Keynes made different assumptions, assumptions that lead to the prediction, impossible in the Neoclassical theory, that the economy can stagnate (get stuck): be in an equilibrium with high unemployment and low income.

(If you are in Econ 4545 and reading these notes for background material, you can stop here.)

Galbraith (1908-1996), a Harvard economist, produced four dozen books and thousands of articles. When I was first studying economics, he was the most famous economist among non-economists. I highly recommend two of his books: The Affluent Society and The New Industrial State.

An example of a simple economic model Your text, Chapter 2, has a simple production model for a one-person economy.
This model in the book has three parts: definitions, assumptions and predictions

The model, as I interpret it, has three variables:

Amount of resources available, Tom Hanks time devoted to production = T Number of coconuts gathered =C Number of fish caught = F The model assumes C and F are endogenous variables what the modeler wants to explain.

The model assumes T is exogenous (the model is not trying to explain how much Tom works) It assumes only two goods can be produced: no cigarettes, booze, or naughty movies, only coconuts and fish

In addition the model assumes time is needed to catch fish and to gather coconuts, and one cannot do both at the same time.

It further assumes that fish caught per hour starts positive, but while remaining positive decreases as Tom allocates more hours to catching fish.

Likewise it assumes the number of coconuts gathered per hour starts positive, but while remaining positive decreases as Tom allocates more of his work hours to gathering coconuts.

What does the model predict?

The model does not predict how many coconuts Tom will gather or how many fish he will catch; rather it predicts the combinations of coconuts and fish that are feasible, and how that will change as Tom works more or less.

What is feasible is represented with a set of points, called a production possibilities set all those combinations of C and F that are feasible.

The boundary of that set is called a production possibilities set.

This graph is a visual representation of the information contained in the assumptions specified on the previous page, plus, sufficient assumptions to imply the numerical amounts.

The assumptions on the previous page together imply the basic shape, but not the specific numbers

Discuss the graph

Again what does this theory predict?

Tom cannot produce outside of the shaded area

What else does it predict?

Does it predict how much will be produced if Tom does not work? This, if Tom does not work then..

What does it predict will happen if Tom increase (or decreases) the amount he works?

What does it predict will happen if Tom continues to work the same amount but catches more fish? Be careful with this question.

What happens if Tom decides to work more or less?

This simple model predicts even more,

As Tom allocates more hours to catching fish, and fewer hours to gathering coconuts, the marginal cost of catching fish in terms of forgone coconuts increases.

An equivalent way of saying this is as Tom allocates more hours to gathering coconuts, and fewer hours to catching fish, the marginal cost of collecting coconuts in terms of forgone fish increases.

To catch another fish requires the sacrifice of some coconuts and each additional fish costs more than the previous one in terms of un-gathered coconuts.

In closing: I have a little trouble with Krugman and Wells identifying the boundary of the production possibilities frontier as efficient.

It is efficient given how much Tom is working, but not necessarily overall efficient.

Requiring Tom to work a certain amount of time might lead to an inefficient amount of work from Toms perspective

It is efficient if Tom chose how much to work.

Efficiency in this world is achieved when Tom is doing the best he can given his constraints (abilities, time, and existing natural resources)

Tom might decide the efficient thing to do is to make coconut beer and drink himself to death, or simply spend all day everyday working on his tan.